Colleges need to recognize this as an existential threat as well as a branding opportunity.
Emma Whitford at Inside Higher Education has a story from Brown University that is hair-raising even by the appalling standards of the current epidemic of students relying on artificial intelligence as a substitute for learning anything.
Brown University economics professor Roberto Serrano found his Welfare Economics and Social Choice Theory class swamped with nearly triple its usual enrollment when he promised to give take-home exams — and then found that the average grade on his midterm was a 96. Suitably alarmed by the uniformity of the grades and other red flags of reliance on AI to write the answers, he challenged the class to prove him wrong by taking an in-class final:
Serrano heard crickets from his students, but 18 of them subsequently dropped the class. Nine students remained enrolled but did not take the final exam. And Serrano said the results proved him right; three students earned a zero, and the average score on the final was 48.6 percent—by far a historic low, he said. Previously, the average final exam score had never dropped below 65 percent. Only a few students scored similarly to how they did on the midterm.
In a follow-up message to students, Serrano announced the midterm was void and the final exam would be worth 80 percent of the students’ final grade. Any student that scored a 40 percent or higher on the final earned a passing grade—previously, he set the passing line at 50 percent or higher. In total, 19 students failed the class.
Whitford includes an eye-popping graphic comparing each student’s (anonymized) midterm and final grades to illustrate the drop-off from midterm to final exam. Unsurprisingly given the culture of today’s elite universities as well as the highly labor-intensive effort required to prove even a single case of AI use (let alone a whole class of them), Serrano has received only tepid support from Brown.
Punishment of individual students guilty of academic dishonesty is important, but realistically, in pursuing it, the colleges and grad schools are using a dish towel to fight a tsunami. They need to start thinking in terms of revamping their entire approach to testing what students are actually learning (so do high schools, but they have different economics and different physical layouts).
Whitford quotes one view from “Tricia Bertram Gallant, director of the Academic Integrity Office and Triton Testing Center at the University of California, San Diego,” which emphasizes shifting the onus of policing misconduct away from professors:
“They don’t get rewarded for dealing with a 60-person case of cheating,” she said. “They are not incentivized to really prevent cheating or to report it when it occurs. . . . Their time is not compensated. If the professor’s a lecturer, and they get paid on the first day of class and stop getting paid on the last day of class, when are they supposed to do it? Or grad students who teach for the summer?” At the same time, institutions can’t sacrifice fairness for efficiency, she said, but there are some steps colleges can take to find a middle ground. Bertram Gallant recommends that academic integrity committees allow students to accept responsibility for cheating via email rather than only during an in-person meeting, and to allow professors to submit complaints against multiple students at once. Ultimately, institutions that care most about upholding academic integrity will dedicate staffing and resources to the matter, she said.
These are all sound points. But at a time when spiraling tuitions driven by bureaucratic bloat are colliding with a demographic cliff in college-age students and a loss of public faith in the value of debt-financed college educations, the absolute last thing colleges need is to find themselves having to hire yet another category of nonteaching administrator — or expanding a new category of discipline that is sure to produce new reams of complaints if the demographics of those who get caught are not in every way proportional to the student body.
Some professors, and many high school teachers, are resorting to going back to what exams were often like 30 or 40 years ago: in-class, handwritten, and closed-book. Even if conducted in rooms without internet access or with phones away, that opens two more hornet’s nests: many students now are typing or texting natives with little experience handwriting anything legible, and the proliferation of extra-time requirements due to various real and claimed learning disabilities makes in-class proctored exams a huge headache to administer and correct. What may be necessary to address at least some of this is to build dedicated test-taking rooms with wired but internet-free low-tech terminals for typing exams and filing them to be collected for grading.
Whatever the mix of prophylactic and investigative tools used to prevent and punish the use of AI, the colleges need to recognize this as an existential threat as well as a branding opportunity. Any college is apt to turn out some people who cheated their way through school, but if a huge proportion of your graduates are simply incapable of doing the work that was required for their diploma, employers are sooner or later going to discount the value of that diploma. By contrast, schools that can instill confidence that they have actually tested their students will find themselves at a competitive advantage.
For years now, most elite colleges have coasted on reputations earned a long time ago, which remain self-reinforcing simply because the school’s brand name shows that it chose incoming students selectively. But with schools chasing a shrinking universe of students, that status quo won’t last forever. The rain is falling hard now. Start building an ark.